Many people might prefer to be master at something, say feature engineering or deep learning model building or algorithms. If one has to master, it will definitely take years. So should one try to be master of somethings or try to be jack of all?
The answer is, in my opinion, first be jack of all, and then try to master of something. The real business problems have low hanging fruits in an area where you might not be the master. For e.g., a problem may require a simple classification ML model (72% or 74% precision doesn’t really make much difference) but it needs to work real time as a software component. If you as a Data Scientist know only how to build a model and specialised at just that, then you are relying on software engineers for deployment processs entirely. On the contrary your skills are not entirely required.